Scribd
Upload
10 views

T Test

Uploaded by

22003744
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
10 views

T Test

Uploaded by

22003744
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 32

T-test

Test of differences
Choose the relevant test
Tests of Difference
• Investigate a hypothesis of two (or more) groups differing with
respect to measures on a variable.
• Most commonly the mean (average)

• E.g. Do males differ from females on importance of appearance?


Example
• How much your appearance is important to you? (1= not at all,.., 10 =
very much) - Appearance
Types of Tests of Difference
One-Sample t-test
• Looks at one variable from one population(sample)

• What you need


• 1 continuous variable (interval or ratio)
• A comparison point (benchmark or KPI)
One-Sample t-test: Example
• RQ: Are average people per household of Students different from
average number of people per household of whole VN (3.5)?

• Hypotheses:
• H0 (null): Household = 3.5
• H1 (alternative): Household ≠ 3.5

• NOTE: Hypotheses could also be one-tail


• H0: Swinburne Household ≤ 3.5
• H1: Swinburne Household > 3.5
One-Sample t-test: Example
One-tail or Two-tailed t-tests
• Our example was two-tailed, i.e. predicted adifference but not
direction (≠3.5)
• This is the default for SPSS To test a one-tailed hypothesis (e.g. >3.5)
you need to divide the Sig. value by 2:

0.014 / 2 = 0.007
One-tail or Two-tail t-tests
▪ One-tail tests are stronger (more power)
▪ i.e. more likely to find a significant result

▪ BUT we need to predict the direction correctly


▪ Remember we reject our hypothesis if the result is opposite (even if that is
statistically significant)
▪ E.g. we predict males score higher than females. If females score statistically
higher then we are wrong (reject alternative hypothesis)
1. One-Sample t-test: Output
• Mean: What was the score in the sample?

• St.Dev: How varied are the scores?

• Test value: What are we comparing to?

• Mean difference: What is the difference between our result and the test value?

• Sig.: Is it significant (<0.05)?


o Divide by 2 if one-tailed
1. One-Sample t-test: Example
According to ABS, in average it takes 52minutes for people to go to work.
•

•According to an interview at Uni, many people claimed that it takes less


than 5 minutes to get to university
1. One-Sample t-test: Example
RQ: Is the average travel time of students to university lowerthan the
o
general population (52 min).

oHypotheses:
▪ H0: Travel time >= 52

▪H1: Travel time < 52


2. Independent Samples t-test
❑ Compares the means (average) for onevariable between two groups
that are independent of each other

❑ What you need:


• 1 Continuous variable (interval or ratio)
• A categorical variable with two independent groups (e.g. gender)
2. Independent Samples t-test: Example
❑ RQ: Does knowledge of Market Research differ by gender?

❑ Hypotheses:
❑ Two-tailed
▪ H0: There is no difference in knowledge across genders
▪ H1: Knowledge differs by gender
❑ One-tailed
▪ H0: Males do not score higher than females (M≤F)
▪ H1:Males score higher than females (M>F)
2. Independent Samples t-test
2. Independent Samples t-test:Example
2. Independent Samples t-test:Output
1. Mean: What was the score for each group?

2. Variance: How varied are the scores in each group?

3. Mean difference: What is the difference between groups?

4. Sig.: Is it significant (<0.05)?


• Divide by 2 if one-tailed
Example
• How much your appearance is important to you? (1= not at all,.., 10 =
very much) – Appearance
2. Independent Samples t-test:Example
• RQ: Does household electricity consumption higher in the city than in
rural areas

• Hypotheses:
• – One-tailed
• H0: Household electricity consumption is lower in the rural than in city areas
or its equal
• H1:Household electricity consumption is higher in the rural than in city areas
3. Paired Samples t-test
• Compares the means (average) for two related variables, e.g.
▪ Same sample pre-and-post
▪ Same sample on two different questions

• What you need:


▪ 2 measurements from same sample
✔ i.e. 2 different rating questions
3. Paired Samples t-test: Example
• RQ: Is learning about social media tracking seen as more important
than learning about analysing data?

• Hypotheses:
• Two-tail
▪ H0: There is no difference in importance
▪ H1: There is a difference in importance
• One-tail
▪ H0: SM is not more important than AD (SM≤AD)
▪ H1: SM is more important than AD (SM>AD)
3. Paired Samples t-test: Example
3. Paired Samples t-test: Example
3. Paired Samples t-test: Output
• 1. Mean: What was the score for each measurement?
• 2. Variance: How varied are the scores of each measurement?
• 3. Mean paired difference: What is the difference between each
measurement?
• 4. Sig.: Is it significant (<0.05)?
• Divide by 2 if one-tailed
• 5. Confidence Interval: What range are we confident in?
Hypothesis:
• It was hypothesised that police trainees can correctly identify more
license plate numbers AFTER memory training than they could
before training.
Which test of difference to use?
• Comparing a score against a pre-set level?
• E.g. Is satisfaction >4?
• One-sample t-test

• Comparing a score across two groups?


• E.g. Is satisfaction different for males and females?
• Independent samples t-test

• Comparing two different ratings?


• E.g. Is the importance of x different from importance of y?
• Paired samples t-test

• Comparing a score across three or more groups?


• E.g. Does satisfaction differ across retirement statuses?
• ANOVA
Which test would you use?
• Is preference for chocolate (1) higher than preference for jelly (1)?
• Paired-samples test
• Is preference for chocolate (1) > 5?
• One-sample t-test
• Is preference for chocolate different for Millenials, Gen X, and Baby
Boomers (3) - nominal scale?
• ANOVA
• Is preference for chocolate different for males and females (2) –
nominal scale?
• Independent samples test
Summary
Different ways to test differences
• Depend on the type of comparison

2. ANOVA: Comparing means across 3+ groups


3. One-sample t-test: Comparing a sample against a set number
4. Independent samples t-test: Comparing means (average) across
two independent groups
5. Paired samples t-test: Comparing means of two related samples
(e.g. pre-post)

You might also like